I wonder if there are anyone else using CFD to model hydraulic valve flow under high pressure (up to 4000 psi). I am currently using Ansys/Flotran and treating this as a turbulent, adiabatic, incompressible problem (steady state as a start, to make it simple), which has very high velocity gradient near the valve opening (metering region) and sometimes associated with cavitation. The objective is to calculate pressure distribution, based on which the hydraulic force on the boundaries (commonly referred to as flow force in hydraulics industry) can be derived.

Questions: 1. Result (pressure) variation or convergence. I have not been able to achieve good results, mostly due to fluctuation of residual pressure for the iterations, since this problem requires very high accuracy (with normallized iteration variation less than 10E-7). Is ther a better way to handle this? 2. Is Flotran suitable for this? Are there other CFD codes which can handle this type of problems better?

Thanks for your time.

As I have said before, the pressure field and the velocity field are decoupled in the incompressible problem. You can solve the velocity field without knowing the pressure at all. So, whether it is 4000psi or 40psi, the velocity field should be the same. But If you are talking about a 4000psi difference in pressure, then that will affect the mass flow rate and thus the Reynolds number of the flow. The Reynolds number of your flow is important because it will determine whether the flow is laminar or turbulent. In the turbulent flow using pressure-based formulation requires that the pressure field does not change very rapidly throughout the flow field. In the flow through valve case, you will have large flow turning around the corner, and possible flow separations.( don't worry about the cavitation for now). So other than the turbulence models used, you need to ask yourself the question: what is likely the real pressure distribution on the wall? Then, from the pressure distribution, try to come up with a mesh wich you think will be able to handle the pressure profile. That's the first step. You do bring up an interesting question: Is it possible to ask a computer code whether it can solve a particular type of problem? The real question is : What is a CFD code? I think a CFD code is a partial listing of a person's way of solving fluid dynamic problems. In other words, the code itself is not capable of solving the problem alone. ( unless you can come up with CFD-AI code which will run by itself. .....not a very good idea! because you will be looking for new jobs.) So, the first step is refine your mesh. The second step is look for flow separation. If there is a flow separation, then the flow can be either steady or un-steady. ( the formulation also will have impact on whether you will pick up steady or un-steady solution.) At this point, you are in the CFD research domain.

Thank you for responding my posting. Additional background info to my original post: pressure (such as 4000 psi referred) was meant differential which varies depending on the valve opening. Model was solved as turbulent flow, based on high Reynolds number, and steady-state to keep it simple to begin with. Mesh effect on results, as is true with virtually all numerical methods, was constantly checked and even someone from ANSYS/Flotran software development group was hired to help developing initial models.

In early 90's, my colleague had difficulties in getting converged solution for a complex 3-D problem using Flotran. There could be numerous reasons for the failure. It is always a good idea to try different CFD codes to see whether you can get similar results. For structure problems, most people use ANSYS . For fluid dynamics problems most people don't use Flotran. I have not used Flotran code myself ( It was based on finite-element method), so, there is not much I can do to make it work. Be open minded in selecting codes, sometimes, it's simply because the problem you are trying to solve is difficult.

Well, it's still not hopeless at all. There are a couple of things you can do with the code you are using ( any code). (1). run laminar flow case, (2). run low Reynolds number flow case, (3). set the pressure differential to a very low number, (4). increase the pressure to check the solution. Since you have said that the mesh is o.k., you need to identify whether the problem is related to the turbulence modeling ( model itself, wall treatment, solution algorithm), or the Reynolds number. It's a good idea to run a low Reynolds number laminar case, say Re=10 up to Re=400. This should cover the practical turbulent flow range without having to depend on the turbulence modeling. A typical Reynolds number to use is 80. In the Re=100 to 400 range, you should be able to see the flow separation pattern, which is very important because it will interact with the down stream boundary conditions.( not every code can handle flow with an open separation bubble which extend over the exit boundary). There is no need to run cases with Reynolds number higher than 400 in the laminar flow range( to simulate the turbulent Reynolds number.) This also will tell you whether the boundary condition or the domain is properly set. If you are lucky, you can then move on to the turbulent flow calculations. There, you may want to use a fully-developed turbulent pipe flow condition .( turbulent everywhere at the inlet, instead of a strange thin boundary layer not properly defined.) When you have uniform inlet condition with thin boundary layer not properly defined, the flow is neraly on the inviscid side. So, why not make the inlet flow completely turbulent, it will make the inlet turbulent Reynolds number lower ( equivalent to low Reynolds number laminar flow). The guideline is: know your problem and know your tools by checking out it under different simulated conditions, especiall for general purpose codes.

Hi Roger,

I am across an article, which seems to describe a problem similar to what you are working on. Take a look at it, it may be of some help.

Paper: Numerical Simulation of Fluid Flow in Poppet Valves Vaughan et. al Proc Instn Mech Engrs vol 206 , pp 119-127

good luck

Anil

Dear Roger,

in our consultancy we are using sucessfully CFX-TASCflow to compute the flow in hydraulic valves.

For example we have modelled the flow in a valve to be steady state, turbulent, isothermal and barotropic. The pressure differences between inlet and outlet were up to 1000 bar which is about 15000 psi. We had no problems in convergence.

Areas of cavitation could be determined by evaluating the pressure field, but no special cavitation model was used up to now.

Ralph P. Mueller

QUESTIONS ON CFD.

I am writing a comprehensive document on ROTODYNAMIC PUMPS [centrifugal,mixed flow and axial flow] and wish to include the current state of the "role of cfd" . As most engineers know, the developments in the pumping field have much empirical content--so where does cfd score in terms of -actual reliable working pumps, Modelling [Scaleup],pit or sump design [inflow],standards,quality.

ken elms

I had the opportunity to look into the design and analysis of centrifugal pump since early 90's. On the surface, it looks like that you can use various type of blade shapes, but in reality, the pump internal flow is highly 3-D, turbulent, separated flow. The off-design conditions also make the flow separation even worse. The volute design also has a great impact on the efficiency of the pump. I had looked into the blade shape design using higher order curves, axi-symmetric flow field, and 3-D passage flow including inlet, etc. using Navier-Stokes solvers. It is very hard to interpret the 3-D results and come up with the positive design change to improve the efficience, unless the original design is relatively poor. In many cases, I had to rely on the try-and-error approach to reshape the blade to get the best performance. So, my feeling is, in the fine tuning of performance, testing is perhaps the best approach in the pump design, mainly because it is not easy to compute the whole internal separated pump flow field ( even at the design condition, there are flow separations ). In many published results ( experimental or calculation), conclusions are conflicting. It is a complex flow field, even though pumps are used everywhere.

Many thanks for your comments, John.Sorry that I used the hydraulic valve application to reach you. I have scoured the forum ,sites and even the pump zone[usa] to build up some know how on this vast topic.Illness about 20 years ago curtailed my career in rotodynamic pumps [Allen`s Ltd,Bedford,UK].Then the systems were no where near computers,let alone CFD.This project I am trying to put together from home is aimed at young students and anyone else interested in this subject.

I must applaud the USA for its availabilty of information and presentation.My education took an upward turn as an 11y/old when an exchange teacher from Indiana taught me.

You have been most helpful and deeply knowledgeable on this subject of cfd and its real applications or drawbacks.

The positive side of CFD is that one can obtain realistic results by just using a computer ( PC or UNIX orkstation ) with either his own code or commercial codes. With todays standard, the resources requirement is minimum. The negative side is that without the guidance of the experimental data and pictures , it is very hard to determine the quality of the CFD results. The practice to include the full geometry and the mesh independent solution so far can not be easily realized for most 3-D problems. As a result, non-converged results are being used to guide the design which can be very dangerous for the future products design. In theory, CFD is an ideal approach ( it is not a tool yet because it requires a lot of human interaction), in practice, we are not sure whether the CFD results are real or not.

Many thanks John for your most helpful comments which more or less say-yes there is a useful role for CFD, but only when the designers of pumps or anything else for that matter know within reason the full limitations of the codes.

My graduating subjects which go back to the 60`s were Mechanical Engineering and then the famous -factor of safety produced margins in material thicknesses that were uneconomical. Nowadays, the opposite seems to apply. If non-converged results are allowed to escape into the design process then future products must be at risk.

However,the cfd role in pumping looks to be still well in the research domains for the reasons you have explained.CRANFIELD UNIVERSITY[Bedford UK] specialises on Turbomachinery as does the von Karman Inst[Europe].The site servers [commercial codes]are also very helpful.

Do you know if the cfd vocabulary notes-Jess A Michelson [nov `95] -adi ,block-structure etc are still current[ie no new definitions of note since]?

Out of interest,I assume you are in the USA.I was taught by a great fellow from Indiana.